Clean up reading level behind the scenes

codeforamerica / posting-pro

2016 NOLA Fellowship project | A job posting builder that guides employers towards creating more effective job postings that are free of implicit bias.

https://posting-pro.herokuapp.com

MIT License

6 stars 3 forks source link

Clean up reading level behind the scenes #12

Closed antislice closed 7 years ago

antislice commented 8 years ago

Consider averaging different scores for a potentially more accurate score and adding more explanation or an explanatory link.

This issue originally created by @antislice as codeforamerica/joblint#22.

[x] sentences miscounted (consistently 2 more than MS word or readability-score.com
[x] newlines aren't counted as an end-of-sentence (inflating grade level)

pacarole commented 7 years ago

Make reading level more robust (maybe have code that remove newlines before reading level is processed). Consider different algorithms.

antislice commented 7 years ago

TextStatistics.js is reporting an incorrect number of sentences, which is affecting the average words per sentence part of the grade level calculation. For the text in the "dietary aide" template, that results in 2.5 grade levels lower than results from readability-score.com and MS word.